1.Aggregating Nested Transformers ⬇️
Although hierarchical structures are popular in recent vision transformers, they require sophisticated designs and massive datasets to work well. In this work, we explore the idea of nesting basic local transformers on non-overlapping image blocks and aggregating them in a hierarchical manner. We find that the block aggregation function plays a critical role in enabling cross-block non-local information communication. This observation leads us to design a simplified architecture with minor code changes upon the original vision transformer and obtains improved performance compared to existing methods. Our empirical results show that the proposed method NesT converges faster and requires much less training data to achieve good generalization. For example, a NesT with 68M parameters trained on ImageNet for 100/300 epochs achieves
$82.3%/83.8%$ accuracy evaluated on$224\times 224$ image size, outperforming previous methods with up to$57%$ parameter reduction. Training a NesT with 6M parameters from scratch on CIFAR10 achieves$96%$ accuracy using a single GPU, setting a new state of the art for vision transformers. Beyond image classification, we extend the key idea to image generation and show NesT leads to a strong decoder that is 8$\times$ faster than previous transformer based generators. Furthermore, we also propose a novel method for visually interpreting the learned model.
2.Sli2Vol: Annotate a 3D Volume from a Single Slice with Self-Supervised Learning ⬇️
The objective of this work is to segment any arbitrary structures of interest (SOI) in 3D volumes by only annotating a single slice, (i.e. semi-automatic 3D segmentation). We show that high accuracy can be achieved by simply propagating the 2D slice segmentation with an affinity matrix between consecutive slices, which can be learnt in a self-supervised manner, namely slice reconstruction. Specifically, we compare the proposed framework, termed as Sli2Vol, with supervised approaches and two other unsupervised/ self-supervised slice registration approaches, on 8 public datasets (both CT and MRI scans), spanning 9 different SOIs. Without any parameter-tuning, the same model achieves superior performance with Dice scores (0-100 scale) of over 80 for most of the benchmarks, including the ones that are unseen during training. Our results show generalizability of the proposed approach across data from different machines and with different SOIs: a major use case of semi-automatic segmentation methods where fully supervised approaches would normally struggle. The source code will be made publicly available at this https URL.
3.Spatio-Contextual Deep Network Based Multimodal Pedestrian Detection For Autonomous Driving ⬇️
Pedestrian Detection is the most critical module of an Autonomous Driving system. Although a camera is commonly used for this purpose, its quality degrades severely in low-light night time driving scenarios. On the other hand, the quality of a thermal camera image remains unaffected in similar conditions. This paper proposes an end-to-end multimodal fusion model for pedestrian detection using RGB and thermal images. Its novel spatio-contextual deep network architecture is capable of exploiting the multimodal input efficiently. It consists of two distinct deformable ResNeXt-50 encoders for feature extraction from the two modalities. Fusion of these two encoded features takes place inside a multimodal feature embedding module (MuFEm) consisting of several groups of a pair of Graph Attention Network and a feature fusion unit. The output of the last feature fusion unit of MuFEm is subsequently passed to two CRFs for their spatial refinement. Further enhancement of the features is achieved by applying channel-wise attention and extraction of contextual information with the help of four RNNs traversing in four different directions. Finally, these feature maps are used by a single-stage decoder to generate the bounding box of each pedestrian and the score map. We have performed extensive experiments of the proposed framework on three publicly available multimodal pedestrian detection benchmark datasets, namely KAIST, CVC-14, and UTokyo. The results on each of them improved the respective state-of-the-art performance. A short video giving an overview of this work along with its qualitative results can be seen at this https URL.
4.Enhance to Read Better: An Improved Generative Adversarial Network for Handwritten Document Image Enhancement ⬇️
Handwritten document images can be highly affected by degradation for different reasons: Paper ageing, daily-life scenarios (wrinkles, dust, etc.), bad scanning process and so on. These artifacts raise many readability issues for current Handwritten Text Recognition (HTR) algorithms and severely devalue their efficiency. In this paper, we propose an end to end architecture based on Generative Adversarial Networks (GANs) to recover the degraded documents into a clean and readable form. Unlike the most well-known document binarization methods, which try to improve the visual quality of the degraded document, the proposed architecture integrates a handwritten text recognizer that promotes the generated document image to be more readable. To the best of our knowledge, this is the first work to use the text information while binarizing handwritten documents. Extensive experiments conducted on degraded Arabic and Latin handwritten documents demonstrate the usefulness of integrating the recognizer within the GAN architecture, which improves both the visual quality and the readability of the degraded document images. Moreover, we outperform the state of the art in H-DIBCO 2018 challenge, after fine tuning our pre-trained model with synthetically degraded Latin handwritten images, on this task.
5.Deep Learning for Weakly-Supervised Object Detection and Object Localization: A Survey ⬇️
Weakly-Supervised Object Detection (WSOD) and Localization (WSOL), i.e., detecting multiple and single instances with bounding boxes in an image using image-level labels, are long-standing and challenging tasks in the CV community. With the success of deep neural networks in object detection, both WSOD and WSOL have received unprecedented attention. Hundreds of WSOD and WSOL methods and numerous techniques have been proposed in the deep learning era. To this end, in this paper, we consider WSOL is a sub-task of WSOD and provide a comprehensive survey of the recent achievements of WSOD. Specifically, we firstly describe the formulation and setting of the WSOD, including the background, challenges, basic framework. Meanwhile, we summarize and analyze all advanced techniques and training tricks for improving detection performance. Then, we introduce the widely-used datasets and evaluation metrics of WSOD. Lastly, we discuss the future directions of WSOD. We believe that these summaries can help pave a way for future research on WSOD and WSOL.
6.Low Resolution Information Also Matters: Learning Multi-Resolution Representations for Person Re-Identification ⬇️
As a prevailing task in video surveillance and forensics field, person re-identification (re-ID) aims to match person images captured from non-overlapped cameras. In unconstrained scenarios, person images often suffer from the resolution mismatch problem, i.e., \emph{Cross-Resolution Person Re-ID}. To overcome this problem, most existing methods restore low resolution (LR) images to high resolution (HR) by super-resolution (SR). However, they only focus on the HR feature extraction and ignore the valid information from original LR images. In this work, we explore the influence of resolutions on feature extraction and develop a novel method for cross-resolution person re-ID called \emph{\textbf{M}ulti-Resolution \textbf{R}epresentations \textbf{J}oint \textbf{L}earning} (\textbf{MRJL}). Our method consists of a Resolution Reconstruction Network (RRN) and a Dual Feature Fusion Network (DFFN). The RRN uses an input image to construct a HR version and a LR version with an encoder and two decoders, while the DFFN adopts a dual-branch structure to generate person representations from multi-resolution images. Comprehensive experiments on five benchmarks verify the superiority of the proposed MRJL over the relevent state-of-the-art methods.
7.Detecting Biological Locomotion in Video: A Computational Approach ⬇️
Animals locomote for various reasons: to search for food, find suitable habitat, pursue prey, escape from predators, or seek a mate. The grand scale of biodiversity contributes to the great locomotory design and mode diversity. Various creatures make use of legs, wings, fins and other means to move through the world. In this report, we refer to the locomotion of general biological species as biolocomotion. We present a computational approach to detect biolocomotion in unprocessed video.
Significantly, the motion exhibited by the body parts of a biological entity to navigate through an environment can be modeled by a combination of an overall positional advance with an overlaid asymmetric oscillatory pattern, a distinctive signature that tends to be absent in non-biological objects in locomotion. We exploit this key trait of positional advance with asymmetric oscillation along with differences in an object's common motion (extrinsic motion) and localized motion of its parts (intrinsic motion) to detect biolocomotion. An algorithm is developed to measure the presence of these traits in tracked objects to determine if they correspond to a biological entity in locomotion. An alternative algorithm, based on generic features combined with learning is assembled out of components from allied areas of investigation, also is presented as a basis of comparison.
A novel biolocomotion dataset encompassing a wide range of moving biological and non-biological objects in natural settings is provided. Also, biolocomotion annotations to an extant camouflage animals dataset are provided. Quantitative results indicate that the proposed algorithm considerably outperforms the alternative approach, supporting the hypothesis that biolocomotion can be detected reliably based on its distinct signature of positional advance with asymmetric oscillation and extrinsic/intrinsic motion dissimilarity.
8.Disentangled Face Attribute Editing via Instance-Aware Latent Space Search ⬇️
Recent works have shown that a rich set of semantic directions exist in the latent space of Generative Adversarial Networks (GANs), which enables various facial attribute editing applications. However, existing methods may suffer poor attribute variation disentanglement, leading to unwanted change of other attributes when altering the desired one. The semantic directions used by existing methods are at attribute level, which are difficult to model complex attribute correlations, especially in the presence of attribute distribution bias in GAN's training set. In this paper, we propose a novel framework (IALS) that performs Instance-Aware Latent-Space Search to find semantic directions for disentangled attribute editing. The instance information is injected by leveraging the supervision from a set of attribute classifiers evaluated on the input images. We further propose a Disentanglement-Transformation (DT) metric to quantify the attribute transformation and disentanglement efficacy and find the optimal control factor between attribute-level and instance-specific directions based on it. Experimental results on both GAN-generated and real-world images collectively show that our method outperforms state-of-the-art methods proposed recently by a wide margin. Code is available at this https URL.
9.Recent Standard Development Activities on Video Coding for Machines ⬇️
In recent years, video data has dominated internet traffic and becomes one of the major data formats. With the emerging 5G and internet of things (IoT) technologies, more and more videos are generated by edge devices, sent across networks, and consumed by machines. The volume of video consumed by machine is exceeding the volume of video consumed by humans. Machine vision tasks include object detection, segmentation, tracking, and other machine-based applications, which are quite different from those for human consumption. On the other hand, due to large volumes of video data, it is essential to compress video before transmission. Thus, efficient video coding for machines (VCM) has become an important topic in academia and industry. In July 2019, the international standardization organization, i.e., MPEG, created an Ad-Hoc group named VCM to study the requirements for potential standardization work. In this paper, we will address the recent development activities in the MPEG VCM group. Specifically, we will first provide an overview of the MPEG VCM group including use cases, requirements, processing pipelines, plan for potential VCM standards, followed by the evaluation framework including machine-vision tasks, dataset, evaluation metrics, and anchor generation. We then introduce technology solutions proposed so far and discuss the recent responses to the Call for Evidence issued by MPEG VCM group.
10.Edge Detection for Satellite Images without Deep Networks ⬇️
Satellite imagery is widely used in many application sectors, including agriculture, navigation, and urban planning. Frequently, satellite imagery involves both large numbers of images as well as high pixel counts, making satellite datasets computationally expensive to analyze. Recent approaches to satellite image analysis have largely emphasized deep learning methods. Though extremely powerful, deep learning has some drawbacks, including the requirement of specialized computing hardware and a high reliance on training data. When dealing with large satellite datasets, the cost of both computational resources and training data annotation may be prohibitive.
11.Predicting invasive ductal carcinoma using a Reinforcement Sample Learning Strategy using Deep Learning ⬇️
Invasive ductal carcinoma is a prevalent, potentially deadly disease associated with a high rate of morbidity and mortality. Its malignancy is the second leading cause of death from cancer in women. The mammogram is an extremely useful resource for mass detection and invasive ductal carcinoma diagnosis. We are proposing a method for Invasive ductal carcinoma that will use convolutional neural networks (CNN) on mammograms to assist radiologists in diagnosing the disease. Due to the varying image clarity and structure of certain mammograms, it is difficult to observe major cancer characteristics such as microcalcification and mass, and it is often difficult to interpret and diagnose these attributes. The aim of this study is to establish a novel method for fully automated feature extraction and classification in invasive ductal carcinoma computer-aided diagnosis (CAD) systems. This article presents a tumor classification algorithm that makes novel use of convolutional neural networks on breast mammogram images to increase feature extraction and training speed. The algorithm makes two contributions.
12.Context-aware Cross-level Fusion Network for Camouflaged Object Detection ⬇️
Camouflaged object detection (COD) is a challenging task due to the low boundary contrast between the object and its surroundings. In addition, the appearance of camouflaged objects varies significantly, e.g., object size and shape, aggravating the difficulties of accurate COD. In this paper, we propose a novel Context-aware Cross-level Fusion Network (C2F-Net) to address the challenging COD task. Specifically, we propose an Attention-induced Cross-level Fusion Module (ACFM) to integrate the multi-level features with informative attention coefficients. The fused features are then fed to the proposed Dual-branch Global Context Module (DGCM), which yields multi-scale feature representations for exploiting rich global context information. In C2F-Net, the two modules are conducted on high-level features using a cascaded manner. Extensive experiments on three widely used benchmark datasets demonstrate that our C2F-Net is an effective COD model and outperforms state-of-the-art models remarkably. Our code is publicly available at: this https URL.
13.KLIEP-based Density Ratio Estimation for Semantically Consistent Synthetic to Real Images Adaptation in Urban Traffic Scenes ⬇️
Synthetic data has been applied in many deep learning based computer vision tasks. Limited performance of algorithms trained solely on synthetic data has been approached with domain adaptation techniques such as the ones based on generative adversarial framework. We demonstrate how adversarial training alone can introduce semantic inconsistencies in translated images. To tackle this issue we propose density prematching strategy using KLIEP-based density ratio estimation procedure. Finally, we show that aforementioned strategy improves quality of translated images of underlying method and their usability for the semantic segmentation task in the context of autonomous driving.
14.Unsupervised Video Summarization via Multi-source Features ⬇️
Video summarization aims at generating a compact yet representative visual summary that conveys the essence of the original video. The advantage of unsupervised approaches is that they do not require human annotations to learn the summarization capability and generalize to a wider range of domains. Previous work relies on the same type of deep features, typically based on a model pre-trained on ImageNet data. Therefore, we propose the incorporation of multiple feature sources with chunk and stride fusion to provide more information about the visual content. For a comprehensive evaluation on the two benchmarks TVSum and SumMe, we compare our method with four state-of-the-art approaches. Two of these approaches were implemented by ourselves to reproduce the reported results. Our evaluation shows that we obtain state-of-the-art results on both datasets, while also highlighting the shortcomings of previous work with regard to the evaluation methodology. Finally, we perform error analysis on videos for the two benchmark datasets to summarize and spot the factors that lead to misclassifications.
15.Pattern Detection in the Activation Space for Identifying Synthesized Content ⬇️
Generative Adversarial Networks (GANs) have recently achieved unprecedented success in photo-realistic image synthesis from low-dimensional random noise. The ability to synthesize high-quality content at a large scale brings potential risks as the generated samples may lead to misinformation that can create severe social, political, health, and business hazards. We propose SubsetGAN to identify generated content by detecting a subset of anomalous node-activations in the inner layers of pre-trained neural networks. These nodes, as a group, maximize a non-parametric measure of divergence away from the expected distribution of activations created from real data. This enable us to identify synthesised images without prior knowledge of their distribution. SubsetGAN efficiently scores subsets of nodes and returns the group of nodes within the pre-trained classifier that contributed to the maximum score. The classifier can be a general fake classifier trained over samples from multiple sources or the discriminator network from different GANs. Our approach shows consistently higher detection power than existing detection methods across several state-of-the-art GANs (PGGAN, StarGAN, and CycleGAN) and over different proportions of generated content.
16.Social-IWSTCNN: A Social Interaction-Weighted Spatio-Temporal Convolutional Neural Network for Pedestrian Trajectory Prediction in Urban Traffic Scenarios ⬇️
Pedestrian trajectory prediction in urban scenarios is essential for automated driving. This task is challenging because the behavior of pedestrians is influenced by both their own history paths and the interactions with others. Previous research modeled these interactions with pooling mechanisms or aggregating with hand-crafted attention weights. In this paper, we present the Social Interaction-Weighted Spatio-Temporal Convolutional Neural Network (Social-IWSTCNN), which includes both the spatial and the temporal features. We propose a novel design, namely the Social Interaction Extractor, to learn the spatial and social interaction features of pedestrians. Most previous works used ETH and UCY datasets which include five scenes but do not cover urban traffic scenarios extensively for training and evaluation. In this paper, we use the recently released large-scale Waymo Open Dataset in urban traffic scenarios, which includes 374 urban training scenes and 76 urban testing scenes to analyze the performance of our proposed algorithm in comparison to the state-of-the-art (SOTA) models. The results show that our algorithm outperforms SOTA algorithms such as Social-LSTM, Social-GAN, and Social-STGCNN on both Average Displacement Error (ADE) and Final Displacement Error (FDE). Furthermore, our Social-IWSTCNN is 54.8 times faster in data pre-processing speed, and 4.7 times faster in total test speed than the current best SOTA algorithm Social-STGCNN.
17.Anticipating human actions by correlating past with the future with Jaccard similarity measures ⬇️
We propose a framework for early action recognition and anticipation by correlating past features with the future using three novel similarity measures called Jaccard vector similarity, Jaccard cross-correlation and Jaccard Frobenius inner product over covariances. Using these combinations of novel losses and using our framework, we obtain state-of-the-art results for early action recognition in UCF101 and JHMDB datasets by obtaining 91.7 % and 83.5 % accuracy respectively for an observation percentage of 20. Similarly, we obtain state-of-the-art results for Epic-Kitchen55 and Breakfast datasets for action anticipation by obtaining 20.35 and 41.8 top-1 accuracy respectively.
18.Unsupervised Part Segmentation through Disentangling Appearance and Shape ⬇️
We study the problem of unsupervised discovery and segmentation of object parts, which, as an intermediate local representation, are capable of finding intrinsic object structure and providing more explainable recognition results. Recent unsupervised methods have greatly relaxed the dependency on annotated data which are costly to obtain, but still rely on additional information such as object segmentation mask or saliency map. To remove such a dependency and further improve the part segmentation performance, we develop a novel approach by disentangling the appearance and shape representations of object parts followed with reconstruction losses without using additional object mask information. To avoid degenerated solutions, a bottleneck block is designed to squeeze and expand the appearance representation, leading to a more effective disentanglement between geometry and appearance. Combined with a self-supervised part classification loss and an improved geometry concentration constraint, we can segment more consistent parts with semantic meanings. Comprehensive experiments on a wide variety of objects such as face, bird, and PASCAL VOC objects demonstrate the effectiveness of the proposed method.
19.Improving Sign Language Translation with Monolingual Data by Sign Back-Translation ⬇️
Despite existing pioneering works on sign language translation (SLT), there is a non-trivial obstacle, i.e., the limited quantity of parallel sign-text data. To tackle this parallel data bottleneck, we propose a sign back-translation (SignBT) approach, which incorporates massive spoken language texts into SLT training. With a text-to-gloss translation model, we first back-translate the monolingual text to its gloss sequence. Then, the paired sign sequence is generated by splicing pieces from an estimated gloss-to-sign bank at the feature level. Finally, the synthetic parallel data serves as a strong supplement for the end-to-end training of the encoder-decoder SLT framework.
To promote the SLT research, we further contribute CSL-Daily, a large-scale continuous SLT dataset. It provides both spoken language translations and gloss-level annotations. The topic revolves around people's daily lives (e.g., travel, shopping, medical care), the most likely SLT application scenario. Extensive experimental results and analysis of SLT methods are reported on CSL-Daily. With the proposed sign back-translation method, we obtain a substantial improvement over previous state-of-the-art SLT methods.
20.Learning to Detect Fortified Areas ⬇️
High resolution data models like grid terrain models made from LiDAR data are a prerequisite for modern day Geographic Information Systems applications. Besides providing the foundation for the very accurate digital terrain models, LiDAR data is also extensively used to classify which parts of the considered surface comprise relevant elements like water, buildings and vegetation. In this paper we consider the problem of classifying which areas of a given surface are fortified by for instance, roads, sidewalks, parking spaces, paved driveways and terraces. We consider using LiDAR data and orthophotos, combined and alone, to show how well the modern machine learning algorithms Gradient Boosted Trees and Convolutional Neural Networks are able to detect fortified areas on large real world data. The LiDAR data features, in particular the intensity feature that measures the signal strength of the return, that we consider in this project are heavily dependent on the actual LiDAR sensor that made the measurement. This is highly problematic, in particular for the generalisation capability of pattern matching algorithms, as this means that data features for test data may be very different from the data the model is trained on. We propose an algorithmic solution to this problem by designing a neural net embedding architecture that transforms data from all the different sensor systems into a new common representation that works as well as if the training data and test data originated from the same sensor. The final algorithm result has an accuracy above 96 percent, and an AUC score above 0.99.
21.How to Calibrate Your Event Camera ⬇️
We propose a generic event camera calibration framework using image reconstruction. Instead of relying on blinking LED patterns or external screens, we show that neural-network-based image reconstruction is well suited for the task of intrinsic and extrinsic calibration of event cameras. The advantage of our proposed approach is that we can use standard calibration patterns that do not rely on active illumination. Furthermore, our approach enables the possibility to perform extrinsic calibration between frame-based and event-based sensors without additional complexity. Both simulation and real-world experiments indicate that calibration through image reconstruction is accurate under common distortion models and a wide variety of distortion parameters
22.Multiple Domain Experts Collaborative Learning: Multi-Source Domain Generalization For Person Re-Identification ⬇️
Recent years have witnessed significant progress in person re-identification (ReID). However, current ReID approaches suffer from considerable performance degradation when the test target domains exhibit different characteristics from the training ones, known as the domain shift problem. To make ReID more practical and generalizable, we formulate person re-identification as a Domain Generalization (DG) problem and propose a novel training framework, named Multiple Domain Experts Collaborative Learning (MD-ExCo). Specifically, the MD-ExCo consists of a universal expert and several domain experts. Each domain expert focuses on learning from a specific domain, and periodically communicates with other domain experts to regulate its learning strategy in the meta-learning manner to avoid overfitting. Besides, the universal expert gathers knowledge from the domain experts, and also provides supervision to them as feedback. Extensive experiments on DG-ReID benchmarks show that our MD-ExCo outperforms the state-of-the-art methods by a large margin, showing its ability to improve the generalization capability of the ReID models.
23.PSGAN++: Robust Detail-Preserving Makeup Transfer and Removal ⬇️
In this paper, we address the makeup transfer and removal tasks simultaneously, which aim to transfer the makeup from a reference image to a source image and remove the makeup from the with-makeup image respectively. Existing methods have achieved much advancement in constrained scenarios, but it is still very challenging for them to transfer makeup between images with large pose and expression differences, or handle makeup details like blush on cheeks or highlight on the nose. In addition, they are hardly able to control the degree of makeup during transferring or to transfer a specified part in the input face. In this work, we propose the PSGAN++, which is capable of performing both detail-preserving makeup transfer and effective makeup removal. For makeup transfer, PSGAN++ uses a Makeup Distill Network to extract makeup information, which is embedded into spatial-aware makeup matrices. We also devise an Attentive Makeup Morphing module that specifies how the makeup in the source image is morphed from the reference image, and a makeup detail loss to supervise the model within the selected makeup detail area. On the other hand, for makeup removal, PSGAN++ applies an Identity Distill Network to embed the identity information from with-makeup images into identity matrices. Finally, the obtained makeup/identity matrices are fed to a Style Transfer Network that is able to edit the feature maps to achieve makeup transfer or removal. To evaluate the effectiveness of our PSGAN++, we collect a Makeup Transfer In the Wild dataset that contains images with diverse poses and expressions and a Makeup Transfer High-Resolution dataset that contains high-resolution images. Experiments demonstrate that PSGAN++ not only achieves state-of-the-art results with fine makeup details even in cases of large pose/expression differences but also can perform partial or degree-controllable makeup transfer.
24.Performance Analysis of a Foreground Segmentation Neural Network Model ⬇️
In recent years the interest in segmentation has been growing, being used in a wide range of applications such as fraud detection, anomaly detection in public health and intrusion detection. We present an ablation study of FgSegNet_v2, analysing its three stages: (i) Encoder, (ii) Feature Pooling Module and (iii) Decoder. The result of this study is a proposal of a variation of the aforementioned method that surpasses state of the art results. Three datasets are used for testing: CDNet2014, SBI2015 and CityScapes. In CDNet2014 we got an overall improvement compared to the state of the art, mainly in the LowFrameRate subset. The presented approach is promising as it produces comparable results with the state of the art (SBI2015 and Cityscapes datasets) in very different conditions, such as different lighting conditions.
25.FINNger -- Applying artificial intelligence to ease math learning for children ⬇️
Kids have an amazing capacity to use modern electronic devices such as tablets, smartphones, etc. This has been incredibly boosted by the ease of access of these devices given the expansion of such devices through the world, reaching even third world countries. Also, it is well known that children tend to have difficulty learning some subjects at pre-school. We as a society focus extensively on alphabetization, but in the end, children end up having differences in another essential area: Mathematics. With this work, we create the basis for an intuitive application that could join the fact that children have a lot of ease when using such technological applications, trying to shrink the gap between a fun and enjoyable activity with something that will improve the children knowledge and ability to understand concepts when in a low age, by using a novel convolutional neural network to achieve so, named FINNger.
26.Style Similarity as Feedback for Product Design ⬇️
Matching and recommending products is beneficial for both customers and companies. With the rapid increase in home goods e-commerce, there is an increasing demand for quantitative methods for providing such recommendations for millions of products. This approach is facilitated largely by online stores such as Amazon and Wayfair, in which the goal is to maximize overall sales. Instead of focusing on overall sales, we take a product design perspective, by employing big-data analysis for determining the design qualities of a highly recommended product. Specifically, we focus on the visual style compatibility of such products. We build off previous work which implemented a style-based similarity metric for thousands of furniture products. Using analysis and visualization, we extract attributes of furniture products that are highly compatible style-wise. We propose a designer in-the-loop workflow that mirrors methods of displaying similar products to consumers browsing e-commerce websites. Our findings are useful when designing new products, since they provide insight regarding what furniture will be strongly compatible across multiple styles, and hence, more likely to be recommended.
27.SB-GCN: Structured BREP Graph Convolutional Network for Automatic Mating of CAD Assemblies ⬇️
Assembly modeling is a core task of computer aided design (CAD), comprising around one third of the work in a CAD workflow. Optimizing this process therefore represents a huge opportunity in the design of a CAD system, but current research of assembly based modeling is not directly applicable to modern CAD systems because it eschews the dominant data structure of modern CAD: parametric boundary representations (BREPs). CAD assembly modeling defines assemblies as a system of pairwise constraints, called mates, between parts, which are defined relative to BREP topology rather than in world coordinates common to existing work. We propose SB-GCN, a representation learning scheme on BREPs that retains the topological structure of parts, and use these learned representations to predict CAD type mates. To train our system, we compiled the first large scale dataset of BREP CAD assemblies, which we are releasing along with benchmark mate prediction tasks. Finally, we demonstrate the compatibility of our model with an existing commercial CAD system by building a tool that assists users in mate creation by suggesting mate completions, with 72.2% accuracy.
28.Learning a Model-Driven Variational Network for Deformable Image Registration ⬇️
Data-driven deep learning approaches to image registration can be less accurate than conventional iterative approaches, especially when training data is limited. To address this whilst retaining the fast inference speed of deep learning, we propose VR-Net, a novel cascaded variational network for unsupervised deformable image registration. Using the variable splitting optimization scheme, we first convert the image registration problem, established in a generic variational framework, into two sub-problems, one with a point-wise, closed-form solution while the other one is a denoising problem. We then propose two neural layers (i.e. warping layer and intensity consistency layer) to model the analytical solution and a residual U-Net to formulate the denoising problem (i.e. generalized denoising layer). Finally, we cascade the warping layer, intensity consistency layer, and generalized denoising layer to form the VR-Net. Extensive experiments on three (two 2D and one 3D) cardiac magnetic resonance imaging datasets show that VR-Net outperforms state-of-the-art deep learning methods on registration accuracy, while maintains the fast inference speed of deep learning and the data-efficiency of variational model.
29.Self-Guided Instance-Aware Network for Depth Completion and Enhancement ⬇️
Depth completion aims at inferring a dense depth image from sparse depth measurement since glossy, transparent or distant surface cannot be scanned properly by the sensor. Most of existing methods directly interpolate the missing depth measurements based on pixel-wise image content and the corresponding neighboring depth values. Consequently, this leads to blurred boundaries or inaccurate structure of object. To address these problems, we propose a novel self-guided instance-aware network (SG-IANet) that: (1) utilize self-guided mechanism to extract instance-level features that is needed for depth restoration, (2) exploit the geometric and context information into network learning to conform to the underlying constraints for edge clarity and structure consistency, (3) regularize the depth estimation and mitigate the impact of noise by instance-aware learning, and (4) train with synthetic data only by domain randomization to bridge the reality gap. Extensive experiments on synthetic and real world dataset demonstrate that our proposed method outperforms previous works. Further ablation studies give more insights into the proposed method and demonstrate the generalization capability of our model.
30.Occlusion Aware Kernel Correlation Filter Tracker using RGB-D ⬇️
Unlike deep learning which requires large training datasets, correlation filter-based trackers like Kernelized Correlation Filter (KCF) uses implicit properties of tracked images (circulant matrices) for training in real-time. Despite their practical application in tracking, a need for a better understanding of the fundamentals associated with KCF in terms of theoretically, mathematically, and experimentally exists. This thesis first details the workings prototype of the tracker and investigates its effectiveness in real-time applications and supporting visualizations. We further address some of the drawbacks of the tracker in cases of occlusions, scale changes, object rotation, out-of-view and model drift with our novel RGB-D Kernel Correlation tracker. We also study the use of particle filters to improve trackers' accuracy. Our results are experimentally evaluated using a) standard dataset and b) real-time using the Microsoft Kinect V2 sensor. We believe this work will set the basis for a better understanding of the effectiveness of kernel-based correlation filter trackers and to further define some of its possible advantages in tracking.
31.AutoReCon: Neural Architecture Search-based Reconstruction for Data-free Compression ⬇️
Data-free compression raises a new challenge because the original training dataset for a pre-trained model to be compressed is not available due to privacy or transmission issues. Thus, a common approach is to compute a reconstructed training dataset before compression. The current reconstruction methods compute the reconstructed training dataset with a generator by exploiting information from the pre-trained model. However, current reconstruction methods focus on extracting more information from the pre-trained model but do not leverage network engineering. This work is the first to consider network engineering as an approach to design the reconstruction method. Specifically, we propose the AutoReCon method, which is a neural architecture search-based reconstruction method. In the proposed AutoReCon method, the generator architecture is designed automatically given the pre-trained model for reconstruction. Experimental results show that using generators discovered by the AutoRecon method always improve the performance of data-free compression.
32.Smile Like You Mean It: Driving Animatronic Robotic Face with Learned Models ⬇️
Ability to generate intelligent and generalizable facial expressions is essential for building human-like social robots. At present, progress in this field is hindered by the fact that each facial expression needs to be programmed by humans. In order to adapt robot behavior in real time to different situations that arise when interacting with human subjects, robots need to be able to train themselves without requiring human labels, as well as make fast action decisions and generalize the acquired knowledge to diverse and new contexts. We addressed this challenge by designing a physical animatronic robotic face with soft skin and by developing a vision-based self-supervised learning framework for facial mimicry. Our algorithm does not require any knowledge of the robot's kinematic model, camera calibration or predefined expression set. By decomposing the learning process into a generative model and an inverse model, our framework can be trained using a single motor babbling dataset. Comprehensive evaluations show that our method enables accurate and diverse face mimicry across diverse human subjects. The project website is at this http URL
33.Towards Transparent Application of Machine Learning in Video Processing ⬇️
Machine learning techniques for more efficient video compression and video enhancement have been developed thanks to breakthroughs in deep learning. The new techniques, considered as an advanced form of Artificial Intelligence (AI), bring previously unforeseen capabilities. However, they typically come in the form of resource-hungry black-boxes (overly complex with little transparency regarding the inner workings). Their application can therefore be unpredictable and generally unreliable for large-scale use (e.g. in live broadcast). The aim of this work is to understand and optimise learned models in video processing applications so systems that incorporate them can be used in a more trustworthy manner. In this context, the presented work introduces principles for simplification of learned models targeting improved transparency in implementing machine learning for video production and distribution applications. These principles are demonstrated on video compression examples, showing how bitrate savings and reduced complexity can be achieved by simplifying relevant deep learning models.
34.On the Advantages of Multiple Stereo Vision Camera Designs for Autonomous Drone Navigation ⬇️
In this work we showcase the design and assessment of the performance of a multi-camera UAV, when coupled with state-of-the-art planning and mapping algorithms for autonomous navigation. The system leverages state-of-the-art receding horizon exploration techniques for Next-Best-View (NBV) planning with 3D and semantic information, provided by a reconfigurable multi stereo camera system. We employ our approaches in an autonomous drone-based inspection task and evaluate them in an autonomous exploration and mapping scenario. We discuss the advantages and limitations of using multi stereo camera flying systems, and the trade-off between number of cameras and mapping performance.
35.Dynamic Probabilistic Pruning: A general framework for hardware-constrained pruning at different granularities ⬇️
Unstructured neural network pruning algorithms have achieved impressive compression rates. However, the resulting - typically irregular - sparse matrices hamper efficient hardware implementations, leading to additional memory usage and complex control logic that diminishes the benefits of unstructured pruning. This has spurred structured coarse-grained pruning solutions that prune entire filters or even layers, enabling efficient implementation at the expense of reduced flexibility. Here we propose a flexible new pruning mechanism that facilitates pruning at different granularities (weights, kernels, filters/feature maps), while retaining efficient memory organization (e.g. pruning exactly k-out-of-n weights for every output neuron, or pruning exactly k-out-of-n kernels for every feature map). We refer to this algorithm as Dynamic Probabilistic Pruning (DPP). DPP leverages the Gumbel-softmax relaxation for differentiable k-out-of-n sampling, facilitating end-to-end optimization. We show that DPP achieves competitive compression rates and classification accuracy when pruning common deep learning models trained on different benchmark datasets for image classification. Relevantly, the non-magnitude-based nature of DPP allows for joint optimization of pruning and weight quantization in order to even further compress the network, which we show as well. Finally, we propose novel information theoretic metrics that show the confidence and pruning diversity of pruning masks within a layer.
36.Blurs Make Results Clearer: Spatial Smoothings to Improve Accuracy, Uncertainty, and Robustness ⬇️
Bayesian neural networks (BNNs) have shown success in the areas of uncertainty estimation and robustness. However, a crucial challenge prohibits their use in practice: Bayesian NNs require a large number of predictions to produce reliable results, leading to a significant increase in computational cost. To alleviate this issue, we propose spatial smoothing, a method that ensembles neighboring feature map points of CNNs. By simply adding a few blur layers to the models, we empirically show that the spatial smoothing improves accuracy, uncertainty estimation, and robustness of BNNs across a whole range of ensemble sizes. In particular, BNNs incorporating the spatial smoothing achieve high predictive performance merely with a handful of ensembles. Moreover, this method also can be applied to canonical deterministic neural networks to improve the performances. A number of evidences suggest that the improvements can be attributed to the smoothing and flattening of the loss landscape. In addition, we provide a fundamental explanation for prior works - namely, global average pooling, pre-activation, and ReLU6 - by addressing to them as special cases of the spatial smoothing. These not only enhance accuracy, but also improve uncertainty estimation and robustness by making the loss landscape smoother in the same manner as the spatial smoothing. The code is available at this https URL.
37.Predict then Interpolate: A Simple Algorithm to Learn Stable Classifiers ⬇️
We propose Predict then Interpolate (PI), a simple algorithm for learning correlations that are stable across environments. The algorithm follows from the intuition that when using a classifier trained on one environment to make predictions on examples from another environment, its mistakes are informative as to which correlations are unstable. In this work, we prove that by interpolating the distributions of the correct predictions and the wrong predictions, we can uncover an oracle distribution where the unstable correlation vanishes. Since the oracle interpolation coefficients are not accessible, we use group distributionally robust optimization to minimize the worst-case risk across all such interpolations. We evaluate our method on both text classification and image classification. Empirical results demonstrate that our algorithm is able to learn robust classifiers (outperforms IRM by 23.85% on synthetic environments and 12.41% on natural environments). Our code and data are available at this https URL.
38.Adversarial robustness against multiple $l_p$ -threat models at the price of one and how to quickly fine-tune robust models to another threat model ⬇️
Adversarial training (AT) in order to achieve adversarial robustness wrt single
$l_p$ -threat models has been discussed extensively. However, for safety-critical systems adversarial robustness should be achieved wrt all$l_p$ -threat models simultaneously. In this paper we develop a simple and efficient training scheme to achieve adversarial robustness against the union of$l_p$ -threat models. Our novel$l_1+l_\infty$ -AT scheme is based on geometric considerations of the different$l_p$ -balls and costs as much as normal adversarial training against a single$l_p$ -threat model. Moreover, we show that using our$l_1+l_\infty$ -AT scheme one can fine-tune with just 3 epochs any$l_p$ -robust model (for$p \in {1,2,\infty}$ ) and achieve multiple norm adversarial robustness. In this way we boost the previous state-of-the-art reported for multiple-norm robustness by more than$6%$ on CIFAR-10 and report up to our knowledge the first ImageNet models with multiple norm robustness. Moreover, we study the general transfer of adversarial robustness between different threat models and in this way boost the previous SOTA$l_1$ -robustness on CIFAR-10 by almost$10%$ .
39.Calibrated prediction in and out-of-domain for state-of-the-art saliency modeling ⬇️
Since 2014 transfer learning has become the key driver for the improvement of spatial saliency prediction; however, with stagnant progress in the last 3-5 years. We conduct a large-scale transfer learning study which tests different ImageNet backbones, always using the same read out architecture and learning protocol adopted from DeepGaze II. By replacing the VGG19 backbone of DeepGaze II with ResNet50 features we improve the performance on saliency prediction from 78% to 85%. However, as we continue to test better ImageNet models as backbones (such as EfficientNetB5) we observe no additional improvement on saliency prediction. By analyzing the backbones further, we find that generalization to other datasets differs substantially, with models being consistently overconfident in their fixation predictions. We show that by combining multiple backbones in a principled manner a good confidence calibration on unseen datasets can be achieved. This yields a significant leap in benchmark performance in and out-of-domain with a 15 percent point improvement over DeepGaze II to 93% on MIT1003, marking a new state of the art on the MIT/Tuebingen Saliency Benchmark in all available metrics (AUC: 88.3%, sAUC: 79.4%, CC: 82.4%).
40.Towards an IMU-based Pen Online Handwriting Recognizer ⬇️
Most online handwriting recognition systems require the use of specific writing surfaces to extract positional data. In this paper we present a online handwriting recognition system for word recognition which is based on inertial measurement units (IMUs) for digitizing text written on paper. This is obtained by means of a sensor-equipped pen that provides acceleration, angular velocity, and magnetic forces streamed via Bluetooth. Our model combines convolutional and bidirectional LSTM networks, and is trained with the Connectionist Temporal Classification loss that allows the interpretation of raw sensor data into words without the need of sequence segmentation. We use a dataset of words collected using multiple sensor-enhanced pens and evaluate our model on distinct test sets of seen and unseen words achieving a character error rate of 17.97% and 17.08%, respectively, without the use of a dictionary or language model
41.Weighing Features of Lung and Heart Regions for Thoracic Disease Classification ⬇️
Chest X-rays are the most commonly available and affordable radiological examination for screening thoracic diseases. According to the domain knowledge of screening chest X-rays, the pathological information usually lay on the lung and heart regions. However, it is costly to acquire region-level annotation in practice, and model training mainly relies on image-level class labels in a weakly supervised manner, which is highly challenging for computer-aided chest X-ray screening. To address this issue, some methods have been proposed recently to identify local regions containing pathological information, which is vital for thoracic disease classification. Inspired by this, we propose a novel deep learning framework to explore discriminative information from lung and heart regions. We design a feature extractor equipped with a multi-scale attention module to learn global attention maps from global images. To exploit disease-specific cues effectively, we locate lung and heart regions containing pathological information by a well-trained pixel-wise segmentation model to generate binarization masks. By introducing element-wise logical AND operator on the learned global attention maps and the binarization masks, we obtain local attention maps in which pixels are
$1$ for lung and heart region and$0$ for other regions. By zeroing features of non-lung and heart regions in attention maps, we can effectively exploit their disease-specific cues in lung and heart regions. Compared to existing methods fusing global and local features, we adopt feature weighting to avoid weakening visual cues unique to lung and heart regions. Evaluated by the benchmark split on the publicly available chest X-ray14 dataset, the comprehensive experiments show that our method achieves superior performance compared to the state-of-the-art methods.
42.Permutation invariance and uncertainty in multitemporal image super-resolution ⬇️
Recent advances have shown how deep neural networks can be extremely effective at super-resolving remote sensing imagery, starting from a multitemporal collection of low-resolution images. However, existing models have neglected the issue of temporal permutation, whereby the temporal ordering of the input images does not carry any relevant information for the super-resolution task and causes such models to be inefficient with the, often scarce, ground truth data that available for training. Thus, models ought not to learn feature extractors that rely on temporal ordering. In this paper, we show how building a model that is fully invariant to temporal permutation significantly improves performance and data efficiency. Moreover, we study how to quantify the uncertainty of the super-resolved image so that the final user is informed on the local quality of the product. We show how uncertainty correlates with temporal variation in the series, and how quantifying it further improves model performance. Experiments on the Proba-V challenge dataset show significant improvements over the state of the art without the need for self-ensembling, as well as improved data efficiency, reaching the performance of the challenge winner with just 25% of the training data.
43.CBANet: Towards Complexity and Bitrate Adaptive Deep Image Compression using a Single Network ⬇️
In this paper, we propose a new deep image compression framework called Complexity and Bitrate Adaptive Network (CBANet), which aims to learn one single network to support variable bitrate coding under different computational complexity constraints. In contrast to the existing state-of-the-art learning based image compression frameworks that only consider the rate-distortion trade-off without introducing any constraint related to the computational complexity, our CBANet considers the trade-off between the rate and distortion under dynamic computational complexity constraints. Specifically, to decode the images with one single decoder under various computational complexity constraints, we propose a new multi-branch complexity adaptive module, in which each branch only takes a small portion of the computational budget of the decoder. The reconstructed images with different visual qualities can be readily generated by using different numbers of branches. Furthermore, to achieve variable bitrate decoding with one single decoder, we propose a bitrate adaptive module to project the representation from a base bitrate to the expected representation at a target bitrate for transmission. Then it will project the transmitted representation at the target bitrate back to that at the base bitrate for the decoding process. The proposed bit adaptive module can significantly reduce the storage requirement for deployment platforms. As a result, our CBANet enables one single codec to support multiple bitrate decoding under various computational complexity constraints. Comprehensive experiments on two benchmark datasets demonstrate the effectiveness of our CBANet for deep image compression.
44.Using the Overlapping Score to Improve Corruption Benchmarks ⬇️
Neural Networks are sensitive to various corruptions that usually occur in real-world applications such as blurs, noises, low-lighting conditions, etc. To estimate the robustness of neural networks to these common corruptions, we generally use a group of modeled corruptions gathered into a benchmark. Unfortunately, no objective criterion exists to determine whether a benchmark is representative of a large diversity of independent corruptions. In this paper, we propose a metric called corruption overlapping score, which can be used to reveal flaws in corruption benchmarks. Two corruptions overlap when the robustnesses of neural networks to these corruptions are correlated. We argue that taking into account overlappings between corruptions can help to improve existing benchmarks or build better ones.
45.What data do we need for training an AV motion planner? ⬇️
We investigate what grade of sensor data is required for training an imitation-learning-based AV planner on human expert demonstration. Machine-learned planners are very hungry for training data, which is usually collected using vehicles equipped with the same sensors used for autonomous operation. This is costly and non-scalable. If cheaper sensors could be used for collection instead, data availability would go up, which is crucial in a field where data volume requirements are large and availability is small. We present experiments using up to 1000 hours worth of expert demonstration and find that training with 10x lower-quality data outperforms 1x AV-grade data in terms of planner performance. The important implication of this is that cheaper sensors can indeed be used. This serves to improve data access and democratize the field of imitation-based motion planning. Alongside this, we perform a sensitivity analysis of planner performance as a function of perception range, field-of-view, accuracy, and data volume, and the reason why lower-quality data still provide good planning results.
46.SimNet: Learning Reactive Self-driving Simulations from Real-world Observations ⬇️
In this work, we present a simple end-to-end trainable machine learning system capable of realistically simulating driving experiences. This can be used for the verification of self-driving system performance without relying on expensive and time-consuming road testing. In particular, we frame the simulation problem as a Markov Process, leveraging deep neural networks to model both state distribution and transition function. These are trainable directly from the existing raw observations without the need for any handcrafting in the form of plant or kinematic models. All that is needed is a dataset of historical traffic episodes. Our formulation allows the system to construct never seen scenes that unfold realistically reacting to the self-driving car's behaviour. We train our system directly from 1,000 hours of driving logs and measure both realism, reactivity of the simulation as the two key properties of the simulation. At the same time, we apply the method to evaluate the performance of a recently proposed state-of-the-art ML planning system trained from human driving logs. We discover this planning system is prone to previously unreported causal confusion issues that are difficult to test by non-reactive simulation. To the best of our knowledge, this is the first work that directly merges highly realistic data-driven simulations with a closed-loop evaluation for self-driving vehicles. We make the data, code, and pre-trained models publicly available to further stimulate simulation development.
47.Graph Self Supervised Learning: the BT, the HSIC, and the VICReg ⬇️
Self-supervised learning and pre-training strategies have developed over the last few years especially for Convolutional Neural Networks (CNNs). Recently application of such methods can also be noticed for Graph Neural Networks (GNNs). In this paper, we have used a graph based self-supervised learning strategy with different loss functions (Barlow Twins[ 7], HSIC[ 4], VICReg[ 1]) which have shown promising results when applied with CNNs previously. We have also proposed a hybrid loss function combining the advantages of VICReg and HSIC and called it as VICRegHSIC. The performance of these aforementioned methods have been compared when applied to two different datasets namely MUTAG and PROTEINS. Moreover, the impact of different batch sizes, projector dimensions and data augmentation strategies have also been explored. The results are preliminary and we will be continuing to explore with other datasets.
48.The Nonlinearity Coefficient -- A Practical Guide to Neural Architecture Design ⬇️
In essence, a neural network is an arbitrary differentiable, parametrized function. Choosing a neural network architecture for any task is as complex as searching the space of those functions. For the last few years, 'neural architecture design' has been largely synonymous with 'neural architecture search' (NAS), i.e. brute-force, large-scale search. NAS has yielded significant gains on practical tasks. However, NAS methods end up searching for a local optimum in architecture space in a small neighborhood around architectures that often go back decades, based on CNN or LSTM.
In this work, we present a different and complementary approach to architecture design, which we term 'zero-shot architecture design' (ZSAD). We develop methods that can predict, without any training, whether an architecture will achieve a relatively high test or training error on a task after training. We then go on to explain the error in terms of the architecture definition itself and develop tools for modifying the architecture based on this explanation. This confers an unprecedented level of control on the deep learning practitioner. They can make informed design decisions before the first line of code is written, even for tasks for which no prior art exists.
Our first major contribution is to show that the 'degree of nonlinearity' of a neural architecture is a key causal driver behind its performance, and a primary aspect of the architecture's model complexity. We introduce the 'nonlinearity coefficient' (NLC), a scalar metric for measuring nonlinearity. Via extensive empirical study, we show that the value of the NLC in the architecture's randomly initialized state before training is a powerful predictor of test error after training and that attaining a right-sized NLC is essential for attaining an optimal test error. The NLC is also conceptually simple, well-defined for any feedforward network, easy and cheap to compute, has extensive theoretical, empirical and conceptual grounding, follows instructively from the architecture definition, and can be easily controlled via our 'nonlinearity normalization' algorithm. We argue that the NLC is the most powerful scalar statistic for architecture design specifically and neural network analysis in general. Our analysis is fueled by mean field theory, which we use to uncover the 'meta-distribution' of layers.
Beyond the NLC, we uncover and flesh out a range of metrics and properties that have a significant explanatory influence on test and training error. We go on to explain the majority of the error variation across a wide range of randomly generated architectures with these metrics and properties. We compile our insights into a practical guide for architecture designers, which we argue can significantly shorten the trial-and-error phase of deep learning deployment.
Our results are grounded in an experimental protocol that exceeds that of the vast majority of other deep learning studies in terms of carefulness and rigor. We study the impact of e.g. dataset, learning rate, floating-point precision, loss function, statistical estimation error and batch inter-dependency on performance and other key properties. We promote research practices that we believe can significantly accelerate progress in architecture design research.